X-Git-Url: https://git.ao2.it/tweeper.git/blobdiff_plain/d4c68dbd5d51d004acfebaa136e56bf017c1c1c1..7097a8ad2ef040bc81a8c5f7ed7cc02e0073eaab:/tweeper.php?ds=sidebyside

diff --git a/tweeper.php b/tweeper.php
index 0c2bdbf..efc0fd6 100644
--- a/tweeper.php
+++ b/tweeper.php
@@ -2,7 +2,7 @@
 /*
  * tweeper - a Twitter to RSS web scraper
  *
- * Copyright (C) 2013-2014  Antonio Ospite <ao2@ao2.it>
+ * Copyright (C) 2013-2015  Antonio Ospite <ao2@ao2.it>
  *
  * This program is free software: you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -18,32 +18,34 @@
  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
  */
 
+require_once 'XML/Serializer.php';
+
 date_default_timezone_set('UTC');
 
 class Tweeper {
 
   private static $USER_AGENT = "Mozilla/5.0 (Windows NT 6.1; rv:22.0) Gecko/20130405 Firefox/22.0";
 
-  public function __construct($stylesheet, $generate_enclosure = FALSE) {
-    $stylesheet_contents = $this->get_contents($stylesheet);
-
-    $xslDoc = new DOMDocument();
-    $xslDoc->loadXML($stylesheet_contents);
-
-    $this->xsltProcessor = new XSLTProcessor();
-    $this->xsltProcessor->registerPHPFunctions();
-    $this->xsltProcessor->setParameter('', 'generateEnclosure', $generate_enclosure);
-    $this->xsltProcessor->importStylesheet($xslDoc);
+  public function __construct($generate_enclosure = FALSE) {
+    $this->generate_enclosure = $generate_enclosure;
   }
 
   public static function epoch_to_gmdate($timestamp)
   {
+    if (!is_numeric($timestamp) || is_nan($timestamp)) {
+      $timestamp = 0;
+    }
+
     return gmdate('D, d M Y H:i:s', $timestamp) . ' GMT';
   }
 
   public static function str_to_gmdate($date)
   {
     $timestamp = strtotime($date);
+    if (FALSE === $timestamp) {
+      $timestamp = 0;
+    }
+
     return Tweeper::epoch_to_gmdate($timestamp);
   }
 
@@ -104,7 +106,11 @@ class Tweeper {
       "video/ogg",
     );
 
-    $url_info = Tweeper::get_info($url);
+    // The RSS specification says that the enclosure element url must be http.
+    // See http://sourceforge.net/p/feedvalidator/bugs/72/
+    $http_url = preg_replace("/^https/", "http", $url);
+
+    $url_info = Tweeper::get_info($http_url);
 
     $supported = in_array($url_info['content_type'], $supported_content_types);
     if (!$supported) {
@@ -152,21 +158,116 @@ class Tweeper {
     error_log($output);
   }
 
-  public function tweep($uri) {
-    $html = Tweeper::get_contents($uri);
+  private function load_stylesheet($host) {
+    $stylesheet = "file://" . __DIR__ . "/rss_converter_" . $host . ".xsl";
+    if (FALSE === file_exists($stylesheet)) {
+      trigger_error("Conversion to RSS not supported for $host ($stylesheet not found)", E_USER_ERROR);
+      return NULL;
+    }
+
+    $stylesheet_contents = $this->get_contents($stylesheet);
+
+    $xslDoc = new DOMDocument();
+    $xslDoc->loadXML($stylesheet_contents);
+
+    $xsltProcessor = new XSLTProcessor();
+    $xsltProcessor->registerPHPFunctions();
+    $xsltProcessor->setParameter('', 'generateEnclosure', $this->generate_enclosure);
+    $xsltProcessor->importStylesheet($xslDoc);
+
+    return $xsltProcessor;
+  }
+
+  private function json_to_xml($html, $json_match_expr, $rootName) {
+    // pre-process, convert json to XML
+    $ret = preg_match($json_match_expr, $html, $matches);
+    if ($ret !== 1) {
+      trigger_error("Cannot match expression: $json_match_expr\n", E_USER_ERROR);
+      return NULL;
+    }
+
+    $data = json_decode($matches[1]);
+    if (!$data) {
+      return NULL;
+    }
+
+    $serializer_options = array (
+      'addDecl' => TRUE,
+      'encoding' => "UTF-8",
+      'indent' => '  ',
+      'rootName' => $rootName,
+    );
 
+    $serializer = new XML_Serializer($serializer_options);
+
+    $status = $serializer->serialize($data);
+    if (PEAR::isError($status)) {
+      trigger_error($status->getMessage(), E_USER_ERROR);
+      return NULL;
+    }
+
+    return $serializer->getSerializedData();
+  }
+
+  private function get_xml_instagram_com($html) {
+    return $this->json_to_xml($html, '/window._sharedData = (.*);/', 'instagram');
+  }
+
+  private function html_to_xml($html, $host) {
     $xmlDoc = new DOMDocument();
 
     // Handle warnings and errors when loading invalid HTML.
     $xml_errors_value = libxml_use_internal_errors(true);
-    $xmlDoc->loadHTML($html);
+
+    // If there is a host-specific method to get the xml data, use it!
+    $get_xml_host_method = 'get_xml_' . str_replace(".", "_", $host);
+    if (method_exists($this, $get_xml_host_method)) {
+      $xml_data = call_user_func_array(array($this, $get_xml_host_method), array($html));
+      $xmlDoc->loadXML($xml_data);
+    } else {
+      $xmlDoc->loadHTML($html);
+    }
+
     foreach (libxml_get_errors() as $xml_error) {
       $this->log_xml_error($xml_error);
     }
     libxml_clear_errors();
     libxml_use_internal_errors($xml_errors_value);
 
-    $output = $this->xsltProcessor->transformToXML($xmlDoc);
+    return $xmlDoc;
+  }
+
+  public function tweep($src_url) {
+    $url = parse_url($src_url);
+    if (FALSE === $url || empty($url["host"])) {
+      trigger_error("Invalid url: $src_url", E_USER_ERROR);
+      return NULL;
+    }
+
+    // Strip the leading www. to be more forgiving on input URLs
+    $host = preg_replace('/^www\./', '', $url["host"]);
+
+    $xsltProcessor = $this->load_stylesheet($host);
+    if (NULL === $xsltProcessor) {
+      return NULL;
+    }
+
+    $html = $this->get_contents($src_url);
+    if (FALSE === $html) {
+      return NULL;
+    }
+
+    $preprocess_html_host_method = 'preprocess_html_' . str_replace(".", "_", $host);
+    if (method_exists($this, $preprocess_html_host_method)) {
+      $html = call_user_func_array(array($this, $preprocess_html_host_method), array($html));
+    }
+
+    $xmlDoc = $this->html_to_xml($html, $host);
+    if (NULL === $xmlDoc) {
+      return NULL;
+    }
+
+    $output = $xsltProcessor->transformToXML($xmlDoc);
 
     if (FALSE === $output) {
       trigger_error('XSL transformation failed.', E_USER_ERROR);
@@ -176,12 +277,18 @@ class Tweeper {
   }
 }
 
+function is_cli()
+{
+  return (substr(php_sapi_name(), 0, 3) == 'cli');
+}
+
 function usage($argv)
 {
-  if (php_sapi_name() != 'cli')
-    $usage = htmlentities("{$_SERVER['SCRIPT_NAME']}?src_url=<src_url>&generate_enclosure=<0|1>");
-  else
+  if (is_cli()) {
     $usage = "{$argv[0]} [-e|-h|--help] <src_url>\n";
+  } else {
+    $usage = htmlentities("{$_SERVER['SCRIPT_NAME']}?src_url=<src_url>&generate_enclosure=<0|1>");
+  }
 
   return "usage: $usage";
 }
@@ -232,12 +339,12 @@ function parse_options_query_string()
 }
 
 
-if (php_sapi_name() != 'cli') {
-  $options = parse_options_query_string();
-  $ERROR_STREAM = fopen('php://output', 'w');
-} else {
+if (is_cli()) {
   $options = parse_options_cli($argv, $argc);
   $ERROR_STREAM = fopen('php://stderr', 'w');
+} else {
+  $options = parse_options_query_string();
+  $ERROR_STREAM = fopen('php://output', 'w');
 }
 
 if (!isset($options['src_url'])) {
@@ -245,17 +352,5 @@ if (!isset($options['src_url'])) {
   exit(1);
 }
 
-$url = parse_url($options['src_url']);
-if (FALSE === $url || empty($url["host"])) {
-  fwrite($ERROR_STREAM, "Invalid url: ${options['src_url']}\n");
-  exit(1);
-}
-
-$stylesheet = "file://" . __DIR__ . "/rss_converter_" . $url["host"] . ".xsl";
-if (FALSE === file_exists($stylesheet)) {
-  fwrite($ERROR_STREAM, "Conversion to RSS not supported: {$url["host"]}\n");
-  exit(1);
-}
-
-$tweeper = new Tweeper($stylesheet, $options['generate_enclosure']);
+$tweeper = new Tweeper($options['generate_enclosure']);
 echo $tweeper->tweep($options['src_url']);